The state of web accessibility in 2026: what 1,564 real websites fail

The sample

The scan database held 2,372 scans when the dataset was generated. To turn that into one observation per website we kept only the most recent completed scan for each site, dropped scans that reached no pages, and dropped single-page scans that reported a perfect 100 with zero issues, because that pattern almost always means the crawler was blocked rather than the site being flawless. 808 scans were excluded on those grounds (137 older duplicates, 247 with no pages, 392 suspicious perfect scores, 32 not completed), leaving 1,564 websites and 4,549 pages.

Sites are mostly commercial: company sites, agencies, e-commerce stores, professional services and local businesses, weighted toward .com domains (1,330 of the sample). Each scan covered the home page and up to four further pages discovered from it, tested with an axe-core based engine at a desktop viewport. The full procedure, filters and known limitations are in the methodology.

Score distribution: a web split in two

Certvo's accessibility score weights each detected issue by severity and normalises for the amount of content scanned, producing a 0 to 100 figure. Across the sample the mean was 65.9 and the median was 72; the gap between them is the first thing the chart shows. Scores do not cluster in the middle. 41.4% of websites scored 80 or above, while 28.0% scored below 50, and comparatively few sit in between.

The split follows from how sites are built. A site on a maintained theme or component library inherits mostly-correct markup and lands in the upper bands with a handful of contrast and link-name issues. A site assembled from page-builder blocks, third-party widgets and hand-written templates accumulates the same few defects on every page and drops into the lower bands. The middle is thin because there is no common way to be half-consistent.

A high score is not a conformance claim. Only 18.8% of sites passed every automated Level AA check, and 48.5% passed every Level A check. A site can score in the 90s while still failing on one criterion, and it can pass every automated check while failing criteria no scanner can evaluate. Read the score as "how much detectable work remains", not as a certificate. The accessibility score explainer covers the formula in detail.

The ten most common failures

The chart ranks axe-core rules by the share of websites with at least one instance. Three findings appear on half or more of all sites; after those the curve drops quickly, which is good news for remediation because fixing the top three, usually in a shared layout or stylesheet, removes most of the issue volume on a typical site.

1. content placed outside any landmark region (62.4% of sites)

The "region" rule fires when page content sits outside every landmark (header, nav, main, footer, aside). Screen-reader users navigate by landmark the way sighted users skim by layout; content outside any landmark is reachable only by reading the whole page linearly. It is a moderate-severity finding on its own, but it is the one that makes every other barrier slower to work around. Affected sites averaged 30.7 instances each.

2. insufficient colour contrast (51.3% of sites)

Colour contrast is the failure most people can see once it is pointed out and almost nobody notices before: pale grey body text, red sale prices on white, light placeholder text. WCAG 1.4.3 asks for 4.5:1 on normal text and 3:1 on large text. It is classified serious because it degrades reading for low-vision users, older users and anyone on a phone in daylight, and it typically appears dozens of times per page because it lives in a shared stylesheet. Affected sites averaged 37.8 instances each, the highest per-site count of any rule in the top ten.

3. links without an accessible name (50.1% of sites)

A link without an accessible name is announced as "link" and nothing else. Icon-only links, image links with empty alt, and "read more" links carrying no context all fail. The user hears a list of forty "link" items and has to guess. This maps to WCAG 2.4.4 and 4.1.2, both Level A, which is why it counts as a Level A failure even on otherwise polished sites. Affected sites averaged 17.5 instances each.

Where a rule has a matching fix recipe on this site, the label above links to it. Each recipe shows the failing pattern, the corrected pattern and the verification steps for the criteria involved.

Failure rates by WCAG 2.1 success criterion

Rules are how a scanner thinks; success criteria are how the law thinks. The European Accessibility Act, through EN 301 549, and the ADA Title II rule both point at WCAG 2.1 Level AA, so this is the table an auditor, a procurement questionnaire or a regulator will read. A website counts as failing a criterion if any detected issue maps to it. 4.1.2 (Name, Role, Value) leads at 67.1% because several high-frequency rules (accessible names for links and buttons, ARIA attribute validity) all map to it.

Severity and conformance level

Not every issue blocks a user. Certvo classifies each finding on the axe-core impact scale. Across the sample, 51.0% of issues were serious and 11.0% were critical, together 62.0% of everything detected. Moderate findings, mostly landmark and heading structure, made up 32.5%, and minor findings 5.6%.

The conformance-level split is the finding that should worry compliance teams most. 71.4% of all issues violate Level A criteria, the minimum every accessibility law starts from, rather than the stricter Level AA. In other words, most of the web's accessibility debt is not a matter of failing to reach an enhanced standard; it is missing alt text, unnamed links and unlabelled buttons that the earliest version of WCAG already prohibited.

Domain groups and site size

The dataset records only the top-level-domain group of each site, never the domain. The groups are unequal in size, so treat the smaller ones as indicative rather than conclusive; a difference of a few points between a group of 1,330 sites and one of 34 is within noise. What the table does show is that no group is close to the AA baseline.

Sites where the crawler found and tested further pages averaged 68.8, against 62.9 for sites where only the home page could be reached. That is partly selection: a home page from which no internal link is discoverable is often one built entirely from scripts, images or an unlabelled menu, which is itself an accessibility failure. It is also a reminder that the home page is usually the best-maintained page on a site, so single-page audits flatter it.

What the data says to do first

The distribution of failures is steep, and steep distributions reward starting at the top. For most sites in this sample the following order removes the majority of detected issues in a day or two of work, because each item lives in one shared place rather than on every page.

1. Wrap the template in landmarks. Add header, nav, main and footer elements (or roles) to the layout once. This clears the most common finding on 62.4% of sites and gives screen-reader users the map they use to skip. See landmarks and skip links.
2. Fix contrast in the design tokens. Change the grey scale and the accent colours in one stylesheet and every page inherits the fix. Use the contrast checker to pick replacements that pass 4.5:1.
3. Name every link and button. Icon links and icon buttons need visible text or an aria-label. This closes both accessible-name rules and moves WCAG 4.1.2 off the top of the failure table. Recipes: empty links, empty buttons.
4. Give every image an alt attribute. Descriptive for informative images, empty for decorative ones. See missing alt text.
5. Re-scan, then test what scanners cannot. Run a free scan to confirm the automated findings are gone, then do a keyboard-only pass and a screen-reader pass of the main task flows. Publish an accessibility statement that says honestly what was tested.

For teams facing the European Accessibility Act the same order applies, with one addition: the statement and the evidence behind it are what a market surveillance authority asks for first, and the penalties for having neither are national and real.

Download the data and how to cite it

The full dataset behind this report is published under a Creative Commons Attribution 4.0 licence. It contains aggregate counts and percentages only.

• Full dataset (JSON) — every table on this page, plus the histogram and methodology counts
• Failure rate per WCAG criterion (CSV)
• Share of sites per axe-core rule (CSV)

Source: Certvo Web Accessibility Report 2026, automated scan of 1,564 websites (2026-03-16 to 2026-09-06). https://certvo.com/research/web-accessibility-report-2026

If you quote a figure, link to this page so readers can check the methodology. If you would like a breakdown that is not published here, for example by industry or platform, contact support@certvo.com.